Data Visualization in R

Robbie Richards

About Myself

  • My research interest is predator-parasite interactions.
  • I use R in my research to:
    • analyze data from the field and lab
    • map out hotspots of disease risk
    • predict species distributions
    • help “future-me” remember what I did
  • I mostly work with small mammals, their predators, and their gastro-intestinal parasites
  • My contact info is robert.richards@uga.edu

Review of past workshops

You should be able to…

  • iterate a calculation
  • simulate populations
  • write a function
  • create an HTML doc with code and commentary
  • write code following best practices
  • calculate summary statistics of a dataset
  • create a figure from data

iteration

Outline

You should be able to…

  • customize figures
  • code a t-test
  • code a correlation
  • code a linear regression

Topics

  1. Approach to visualizations
  2. ggplot review

  3. Plotting challenge

    Break

  4. Inferential statistics

  5. Wrap Up

1. Approach to visualization

Visualizations are useful for data exploration and presentation.

Characteristics of data exploration visualizations

  • complex
  • minimally annotated
  • potentially pattern-less

should be avoided for presentation visualizations.

Presentation visualizations

should guide the reader through the research.

Visualization path

Presentation visualizations

We've spent a bit of time making figures for exploration, but haven't touched on presentation quality visualizations.

What are some characteristics of a good presentation visualization?

2. Review of ggplot

plots are divided into 3 fundamental parts

plot = data + Aesthetics + geometry

Minimium arguements

Using the ggplot2 cheat sheet

The cheat sheet reads like a choose your adventure book.

  1. Based on your data, go to the One, Two or Three variables section.

Using the ggplot2 cheat sheet

The cheat sheet reads like a choose your adventure book.

  1. Based on your data, go to the One, Two or Three variables section.
  2. If Two Variables decided if x and y are continuous.

Using the ggplot2 cheat sheet

The cheat sheet reads like a choose your adventure book.

  1. Based on your data, go to the One, Two or Three variables section.
  2. If Two Variables decided if x and y are continuous.
  3. Start you ggplot based on the example given under the heading

Using the ggplot2 cheat sheet

The cheat sheet reads like a choose your adventure book.

  1. Based on your data, go to the One, Two or Three variables section.
  2. If Two Variables decided if x and y are continuous.
  3. Start you ggplot based on the example given under the heading
  4. Add geom_function and additional aesthetic mapping as needed

Using the ggplot2 cheat sheet

The cheat sheet reads like a choose your adventure book.

  1. Based on your data, go to the One, Two or Three variables section.
  2. If Two Variables decided if x and y are continuous.
  3. Start you ggplot based on the example given under the heading
  4. Add geom_function and additional aesthetic mapping as needed
  5. Repeat last step until finished.

3. Plotting challenge

Often, you'll have an idea of the visualization you'd like to make.

But how to get there?

practice (and google)

3. Plotting challenge

For the visualization exercises you'll work with the people sitting in your row.

  • Open W4_Exercise\DataViz_X.html with the number corresponding to your row.
  • Complete the 2 challenges as a group. (this is a chance to practice starting a new project)
    • Every group has the same 1\( ^{st} \) challenge, but different 2\( ^{nd} \) challenge
    • When you're finished, practice explaining your Challenge 2 code.
  • We'll shuffle groups and explain the solutions to eachother.

If you finish early, add more details to the plots or try exporting as a .png.

3. Plotting challenge

What were some of the interesting things you learned?

Break

When you come back we'll be switching gears to inferential statistics

4. Inferential statistics

is the branch of statistics that deals with hypothesis tests.

R is much better for inferential statistics than other software

why?

Strength of R for statistics

R is much better for inferential statistics than other software

  • many of the common statistical analyses already exist as part of libraries
  • code based analysis means modifications to analysis are recorded
  • …

We want to to ask:


Question

You are tasked with determining if butterfly infection prevalence is related to the region in which the butterflies were sampled and whether there are significant relationships with urbanization metrics. There are many approaches that you could take, but start with the basic inferential statistics to ask:

  1. Is the average monarch infection prevalence different between regions?
  2. Is the prevalence associated with urbanization metrics?
  3. Are any of the urbanization metrics correlated with each other?


Test

  1. T-tests t.test()
  2. Linear regressions lm()
  3. Correlation cor.test()

Start by opening W4_Exercise\W4_Exercise.Rproj

We want to compare regions to ask:


Test

  1. T-tests t.test()
  2. Linear regressions lm()
  3. Correlation cor.test()


Result

  1. the means differ

We want to compare regions to ask:


Test

  1. T-tests t.test()
  2. Linear regressions lm()
  3. Correlation cor.test()


Result

  1. the means differ
  2. some are strongly, positively correlated

We want to compare regions to ask:


Test

  1. T-tests t.test()
  2. Linear regressions lm()
  3. Correlation cor.test()


Result

  1. the means differ
  2. relationships between infection and urbanization
  3. some are strongly, positively correlated

5. Wrap Up: We can ...

  • calculate summary statistics of a dataset
  • create a figure from data
  • create an HTML doc with code and commentary
  • write code following best practices
  • write a function
  • iterate a calculation
  • simulate populations
  • customize figures
  • code basic inferential statistics

iteration